DTIC ADA280966: Computer Aided Engineering (CAE) Overview

AD-E402 647

Rsi|Kjrl AHFSD<TR-93045

* 2.. •• i --^4

iD£D EMGINEEBiMG (CAE) OVERVIEW

Hagai G. Maira

OTIC

ELECTE
JUN 2 9 1994

2

June 1994

J.S. ABMY ARMAMENT RESEARCH, DEVELOPMENT AND

ENGINEERING CENTER

Fire Support Arrmments Center
^atinny Arsenal, Jersey

Approved lor public release; distribution is unlimited.

,p

i98l0

94 6 28 163

The views. opixUons, and/or findings contained in this
report are those of the authorsts) and should not be
construed as an official Department of the Amqr posi¬
tion. policy, or decision, unless so designated by other
documentation.

The citation m this report of the names of commercial
firms or commerclalty available products or services does
not constitute ofilclal endorsement by or approval of
the U.S. Government.

Destroy this report when no longer needed by any method
that prevent disclosure of Its contents or recon¬
struction of the document. Do not return to the origi¬
nator.

REPORT DOCUMENTATION PAGE

Foim Approved OM8 No. 07044188

1

PubUc reporting burden (or ttiis coUection of information is estimaled to average 1 hour per response, including the lime for reviewing
Insiruclioitt, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of
information. Send comments regarding this burden estimale or any other aspect of this collection of information, including suggestions for
reducing this burden, to Washington Headquarters Services. Oksctorate tor Information Operation and reports. 1215 Jefferson Davis
Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-
0188), Washington. DC 20503.

1 . AGENCY USE OM.Y (leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED

June 1994

4. TITLE AND SUBTITLE

(X)MPUTER AIDED ENGINEERING (CAE) OVERVIEW

5. FUNDING NUMBERS

6. AUTHOR(S)

Hegel G. Neira

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESSES(S)

ARDEC, FSAC

Precision Munitions Division
(SMCAR-FSP-E)

Picatinny Arsenal, NJ 07806-5000

8. PERFORMING ORGANIZATION

REPORT NUMBER

Technical Report
ARFSD-TR-93045

9.SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(S)

ARDEC, IMD

STINFO Br. (SMCAR-IMI-I)

Picatinny Arsenal, NJ 07806-5000

10. SPONSORINGJMONITORING

AGENCY REPORT NUMBER

11. SUPPLEMENTARY NOTES

12a. DISTRIBUTION/AVAILABIUTY STATEMENT

Approved for public release; distribution is unlimited.

12b. DISTRIBUTION CODE

13. ABSTRACT (Maximum 200 words)

Computer aided engineering (CAE) refers to a collection of software and hardware tools integrated into a system (a
computer) that is providing the circuit designer and circuit troubleshooter with step-by-step assistance during each
phase of the design and analysis cycle, as well as during development, documentation, and maintenance. Under
the CAE umbrella a number of commonly called "automated design tools," which are the software components of
CAE, are revolutionizing and transforming engineering environments from the "hands-on" way of conducting
business into a virtual or simulated "hands-on" mode of operating; and are having a tremendous impact throughout
all engineering disciplines. They have not yet displaced breadboarding and other methods of developing circuit
boards yet but are making their presence known to the point of being totally necessary in the design of certain
devices. It is the intention of this report to promote the use of these tools in the government by providing en¬
gineering management with an overview of the hardware and software products available for electronic simulation,
while covering trends, new technologies, and costs.

SUBJECT TERMS

Schematic capture Electronic simulation Spice Sensitivity Analysis

Library models Workstations

IS. NUMBER OF PAGES 21

16. PRICE CODE

17. SECURITY CLASSIFICATION
OF REPORT

UNCLASSIFIED

18. SECURITY CLASSIFICATION
OF THIS PAGE

UNCLASSIFIED

19. SECURfTY CLASSIFICATION
OF ABSTRACT

UNCLASSIFIED

20. UMITATION OF ABSTRACT

SAR

NSN 7540-01 280-5500

Standard Form 208 (Rev. 2-89)
Prescrired by ANSI Std. Z39-18
298-102

CONTENTS

Page

introduction 1

CAE at Work 1

Schematic Capture 1

Electronic Design and Analysis 1

Advanced Computer Technology 6

New Personal Computers 6

Workstations 8

Responsible Technology 8

RISC Architecture 8

Justifying and Purchasing 1 0

Systems Costs 1 0

Real Benefits 1 1

The Purchase Decision 1 2

Conclusions 1 3

Distribution List 1 5

Accesion For

NTIS CRA&I
OTIC TAB
Unannounced
Justification

8

□

By _

Dist. ibution /

Availability Codes

Dist

m

Avail and/or
Special

INTRODUCTION

In Computer Aided Engineering’s (CAE) earlier days, only selected organizations
in large companies and government could afford the systems required to automate,
design, and/or manufacture an end-item. The majority of midsize and small or¬
ganizations were shut out from taking advantage of technologies because of exor¬
bitant costs and lack of qualified personnel to help them undertake such transition.
Breakthroughs in integrated circuit (1C) technology and programming have now
opened up the doors of CAE to more modest companies and applications, even to one-
man firms and hobby electronics enthusiasts.

CAE At Work

At work, the simulation of an electronic circuit in a CAE system begins with the
schematic capture of a circuit, where a complete description of the circuit in computer
language is created (often times called a ‘‘net-list”), which is the data used to analyze,
troubleshoot, and print or plot the results.

Schematic Capture

Working in 2-D, designers create circuits using lines and component models.
Rather than redraw a component, a circuit or a section of circuitry used in other areas
can easily be duplicated using the copy and grouping function.

When a CAE package is integrated in an environment where parameters and
blueprints are constantly changing to reflect new product capabilities and
improvements, modifications are quickly made without having to recreate the entire
drawing.

These and other techniques accelerate the analysis of any design which could
include a large number of subassemblies. The experience in many companies is that
productivity had doubled since the systems were installed (ref 1). In the past, en¬
gineers worked with huge drawings that made the analysis of a particular circuit area
very cumbersome.

Electronic Design and Analysis

Design engineers using a CAE or a computer aided design (CAD) system don’t
have to depend on a drafting technician to present a layout of the most current design,
since the data base is automatically updated every time a circuit modification is
performed.

Applications ranging from 10 to 100. to 1000 transistor boards can be simulated
In a CAE system for a personal computer (PC), where the number of transistors that
can be handled is constantly increasing as prices of more powerful PC drop. The limit
is 10,000 discrete transistors at a time. 1991. Although there are other discrete
components that are included in analog and digital libraries, the number of transistors
that an electronic software package can handle is a figure-of-merit, and is something
to look at when choosing a CAE package.

Working with Simuiation Packages

A simulation package can not as yet replace the engineer. It can compute
the voltage, the current, and other electric parameters at virtually any point in the
circuit, but it can’t communicate to you exactly what to do with them or how to achieve
a certain response from the circuit. A CAE package is intended as a design aid and
will do only what it is asked. The right questions have to be asked before every
characteristic of the circuit is known. For example, in power dissipation if you try to
dissipate 2 W into an 1/8 W resistor, the simuiation program probably would not raise a
flag, but will give you a value if a query is made.

This example is not 100% right. As a matter of fact, a package called Smoke
Alarm®i will now raise a flag (in the form of a puff of smoke) on the screen next to the
component in trouble, to indicate that it has exceeded its ratings. Thereby simulating
the actual operating constraints of the circuit.

An electronic simulation program will also warn of simple problems such as
short circuits and other extraordinary conditions, that often times are the culprit in
malfunctioning.

More Simulation Less Breadboarding

in a typical design cycle, breadboarding (physical wiring of a circuit in a test
board) has to be kept to a minimum in order to produce a cost effective board, which in
this case is the product. The optimal process is to use simulation at various times
throughout the development process (trying different formats) to achieve a reliable,
producible design that meets specifications. Experienced designers advise the use of
some level of circuit simulation in the early design stages to learn more about the
circuit before breadboarding. For the case of high bandwidth components, or high
frequency, the introduction of parasitic inductance and capacitance usually requires
the construction of production quality prototypes. Simulation will reduce the number of
time consuming breadboarding iterations required to complete a design.

^Cadence Design Systems, San Jose, CA.

2

The consensus among engineers using CAE packages is that breadboar¬
ding still gives a closer representation of the final circuit for low frequency designs, but
not for high frequency designs. They also agree that simulation works well in some
cases but not for others. For example, simulation for designs using complimentary
metal oxide semiconductor (CMOS) provides a good correlation to the hardware, but
bipolar designs do not offer such a close match.

Spice Simulation. Most analog simulators use Spice^ or a derivative.
The newer derivatives are constantly improving the shortcomings in equation conver¬
gence that the original had. by modifying Spice’s model equations.

Spice simulates a device using both modeled equations and modeled
parameters. The first describes device types and technologies [CMOS, field effect
transistor, transistor, transistor logic, etc.], and are part of the simulation software and
the user typically cannot modify them. The second one models specific device types
within a technology (from data books e.g. Sm. im. etc.). THe user can alter these.

Mixed-mode Simulation. Mixed-mode simulation (ref 2) is on the rise,
with more than a few vendors now offering it.

Because even systems that are mostly digital systems have to sooner or
later interface in the real world (the analog world), both analog and digital simulation
capabilities are needed to simulate an entire system. It is a big plus to have the two
simulation packages integrated into one. where simulation by parts is no longer the
only means to accomplish a circuit simulation.

Mixed-mode simulation is espedaily important for circuit designs that have
feedback paths linking analog and digital circuits. For example, motor or engine
controls, where analog sense or drive circuits and the control equations are implemen¬
ted digitally, would be difficult to simulate wnthout mixed-signal simulation. Breaking
up the analog and digital sections does not result in an effective simulation of the
control hoop.

Mixed-mode simulators (ref 2) typically comprise separate analog and digital
simulators that run in unison. These separate simulators need the net list split into
analog and digital sections before they can run the simulation. A good integration of
the two simulators will handle the net list split internally. The separate analog and
digital simulators operate independently except when an analog input to a digital
device crosses a threshold or a digital input to an analog device changes state.

ZBerkeley Spice created in May 1975 by professor L. Nagel.

3

other Simulation Packages. Simulation packages dedicated to specific
design specialties are becoming widely available, and should ideally be compatible
with other electronic design automation (EDA) software.

Temperature-stress analysis packages are becoming popular in en¬
gineering environments because of valuable information about actual working thermal
distributions, in military applications, where working temperatures are usually higher
than for commercial applications, it is usually imperative to support a design with
considerable analysis, and to show that the circuits will work over a range of
temperatures.

EEsof and Compact Software are two simulation vendors that have developed
tools dedicated to radio frequency (RF) and microwave design. These packages have
extensive military applications and are capable of covering most frequency ranges.

Analogy’s SABER is a simulation package that is more than just an
electronic simulator because it can use both standard electronic device models and
standard mathematical equations to define elements (electronic, mechanical, and
chemical) of a complete system.

Monte Carlo and Sensitivity Analysis. Monte Carlo analysis has the
ability to simulate process-variations (fluctuation in the manufacturing yield) effects
and although an experienced designer will design a circuit to minimize these effects,
the simulation of such variations can result in a better design evaluation.

Sensitivity analysis is used to tighten the performance spread of a circuit.
This simulation package will help to determine which part variations have the most
effect on a circuit, or how sensitive the circuit performance is to the change in com¬
ponent value. Sensitivity analysis lists components starting with those that have the
most effect on the circuit, and calculates the worst-case outcomes.

In system-level circuit design, Monte Carlo and sensitivity analysis will
graphically indicate which components are being excessively stressed; and in
addition, it will list in descending order the percentage of time that each component is
used in the circuit during normal operating conditions. This knowledge will allow
relaxed performance specs and still meet overall performance. It would also prompt
the use of premium parts only where they are absolutely needed to keep cost down
and increase reliability in the long run.

Library Models

A simulation package is limited only by the accuracy of the models that are
used. In the early stages of design, a simulation package enables the user to view,
understand, and analyze a circuit’s response as the circuit is created. The use of
these model libraries does not immune anyone from related modeling problems. A

4

few things need to be kept in mind when using these iibraries, i.e., if the necessary
modei parameters are not avaiiable in the library, they will have to be filled in from
data books and parametric analyzers (programs that help find the optimum parameter
for the given application of the circuit). Also, when a part is modeled for a specific
application, factors can be included that are important to that modei. For example, if a
quick response is desired from a threshold detector or an amplifier, the slew time
becomes a factor that should be closely looked at and modified if necessary in the
library models. When simulation vendors develop parts libraries, they have to satisfy
everyone’s needs, which can result in models that are a compromise. A model that
oversimplifies an important effect being sought will cause trouble. On the other hand,
a more detailed model will always make the simulation run slower. In addition, there
are some variations in the quality of models available, so an assessment of the
limitations of the models to be used is needed.

Component models probably present the biggest problem for most users. In
the Military, were the critical components are custom made for the application or
project at hand; the biggest problem is the modeling of components or availability of
library parts. This fact alone can render big complex simulation packages to no more
than a “bali park figure getter.” The solution to this problem is to contract it out to
simulator vendors or companies that are dedicated to the modeling of electronic
devices, or to create your own models using the simulator’s features in house. The
latter needs specialized training investment, but is the most efficient, cost wise, and
rather applicable when dealing with confidential or secret parts.

Model needs vary greatly from user to user; and no one model can satisfy
every application, because a give device can be modeled at several different levels of
complexity. A transistor modei may have less than a dozen or as many as several
dozen parameters (a graduate level electronics test book (ref 3) lists 23 parameters for
a typical bipolar transistor).

Valid Logic Systems (now Cadence Design Systems), for example,
publishes data books along with its models that show characteristic device curves.
The books detail what effects have been factored in and those that have not. The
system permits making parametric changes quickly.

As new devices are introduced, simulation users must either obtain or
develop models of the devices. The best solution is to have the device vendors
provide models of new parts as they are introduced. Needless to say, device vendors
are already doing just that since it makes their parts more attractive to the even larger
community of designers that use simulation as a designing tool.

5

ADVANCED COMPUTER TECHNOLOGY

The technology that promised to change the engineering industry as we know it,
is delivering. Constant new versions and new generations of hardware and software,
specially in the CAD arena, are commonplace and affordable to virtuai.y all en¬
gineering organizations.

New Personal Computers

PCs are no longer a weak platform. The days when PCs were used when nothing
else could be afforded are history. The latest and more powerful PCs offer the
processing power, graphics, and memory that are required to tackle a good share of
the CAD and CAE software in the market today.

More Powerful Hardware

After the introduction of PCs in the mid-seventies, the computer industry has
seen many changes. Today’s PCs have the power of mainframes of not too long ago.
These days PCs are rather inexpensive if compared to the cost of . . . let’s say, sending
someone on travel for a few days or buying a conference chair. PCs lend themselves
well to expansion and upgrades. More memory can be easily added. Faster and
higher resolution graphics can also be installed. Lately, some manufacturers have
developed adapted versions of newer generation microprocessors that can be
mounted in old computers; more specifically, in old “mother-boards” (main electronic
board that basically contains the bus system and the microprocessor), making the
upgrade alternative even more cost effective.

Expanded memory (more RAM readily available for the operating system’s
use during processing), coupled with newer and better operating systems like Unix3
and operating system/2 (OS/2)4, that are capable of multitasking (the ability to run
more than one program or application at the time), are making PC-based tools the
engineer’s workhorse for even more complex design tasks.

According to Byte magazine (ref 4), there are more engineering groups that
have bought tools that run on an IBM PC and its clones than any other platform.

Automation tools for PCs are relatively inexpensive when compared to tools
that run on the Sun, Hewlett Packard (HP), or Silicon Graphics station; and automation
tools for PCs would certainly provide a training vehicle before investing far larger sums
for workstation-based tools.

30perating system developed at Bell Labs in the early 70’s capable of addressing
more memory at a time than DOS, and based on a more efficient ‘higher language,
“C”’.

^Similar to Unix OS, but developed at IBM in the early 80’s.

6

Vendors of workstation-based engineering design automation (EDA) tools
are looking to these new powerful PC as the future platform for their software package,
and as a way of reaching more and more engineers. PSpice®s and MatLab®^ both
have versions for PCs and workstations running on Unix.

PC-based tools can handle the same types of design tasks as workstation-
based EDA tools but on a lower scale. They can handle schematic capture or drafting,
logic and analog simulation, signal processing. 1C design, and PCB design.

Newer Operating Systems

DOS can only access 640 Kbytes of memory. Because of this memory
addressing limitation, the so called DOS extenders were created. They allow software
developers to bypass the memory limitation and make full use of the 32 bit
microprocessor machines now widely available. The latest generation of operating
systems like OS/2 and Unix, also Windows NT, for new technology that is due out at
the end of 1 993, provides multitasking and overcomes the serious limitations of DOS.
It removes the 640 Kbytes barrier, provides 16 Mbytes of addressable memory and 48
Mbytes of virtual memory (memory that resides in the hard drive but is looked upon by
the microprocessor as RAM memory available for the operating system to be used
when application programs.

Powerful Engines

Central processing units (CPU) are indeed the computer’s engine, and
already a 32-bit CPU based machine running at about 66 MHz is currently available
for less than $15,000.00. These systems are more than 25 times faster than the
original PC of the early seventies. Intel and Motorola are at the forefront of the
microprocessor race. They are constantly announcing a new generation of
microprocessors every year, breaking new grounds in transistor density and proces¬
sing speed. The Intel’s 80486 or i486, and Motorola’s 68040 contain over one million
transistors. The next generation of these chips is rumored to contain about three times
as many transistors.

Although Intel-based computers or IBM-compatibles have a bigger share of
the PC hardware market, and of the software market, Apple computers are promising a
good fight. Because of great reviews received about newer micros from Motorola,
Apple computers (which use these new micros in its systems) are poised to stand their
own ground in the competitive PC arena bringing higher powered and cheaper units
in the future.

5A mixed-mode electronic simulation package by MicroSim Corp.

6A mathematical and signal processing program by Math Works Incorporated.

7

Workstations

Workstations have become the platform of choice of CAE users. The machines’
multitasking Unix operating system, huge memories, built-in networking and fast
graphics, along with usually large high definition (a 1024 by 760 pixels is considered a
high definition monitor) monitors make them ideal for CAE applications.

Workstations are known for their ability to run large programs, with great speeds
(in the range of tens of millions of instructions per second), having virtually all
manufacturers offering network and interfaces to numerous computer environments.
These machines can also be enhances by adding more memory, dedicated graphics,
or floating point processors.

Low-end workstations in 1990 could be acquired at a price of about $3,000 to
$10,0007. These machines are often monochrome, and are used for drafting, some
electronic design, and various graphic applications. A step up from that are machines
that provide color and have faster processors equipped with at least 32 Mbytes of
RAM. These usually handle 3D modeling and analysis of all types. They sell for
$10,000 to $30,000, but the prices are less than a year ago and they keep coming
down due to stiff competition among workstation vendors as well as competition from
high-end PC. The drop in unit cost is perhaps the most significant development in
workstations as their performance continues to increase.

Responsible Technologies

What has made these high powered computers available at low costs are two
developments: a reduced instruction set computing (RISC) architecture and an
emphasis on user interface graphics for which new programming techniques and
faster hardware are responsible.

RISC Architecture

The developers of the RISC architecture derived the fundamental concepts of this
design philosophy after analyzing millions of lines of existing computer code. This
analysis testified to the fact that most of the software in use at the time (late seventies)
did not employ the then de facto computer architecture complex instruction set
computing (CISC) in a very efficient manner.

The term reduced instruction set computing does not manifest its true characteris¬
tics. The goal of a RISC architecture is not to reduce the number of executable
instruction in Its instruction set but rather to speed up the processing by making the

7AII monies from FY 90.

8

compilers (original to a particular computer hardware, it is a program in read only
memory that is the intermediary between software and hardware) and processors
match. When developing RISC architecture developers had at their disposal, due to
advances in the field of electronics, wider bandwidths at higher frequencies and faster
memories that, along with the newer 32 bit CPU, allowed them to cram more infor¬
mation in every instruction with the objective to accomplish an instruction of a fixed
length (32 bits) with every cycle. These days a typical RISC architecture machine
would usually include: a single-cycle instruction execution, a fixed length instruction,
larger register sets, and would support the so called "high-level” languages like Unix.

In a CISC processor there is a control program called microcode, which interprets
and supervises the incremental instruction execution, playing the role of "middle man”
between the compiler and the CPU. A RISC processor has no microcode. The
machine’s instructions, now crammed with more information, become the microcode.
Consequently, the theoretical achievement is one instruction per clock cycle; however,
no pP has yet attained that goal. It has achieved an average of close to a one
instruction per cycle.

X Window System

As workstations become more popular users are demanding that they be as
easy to use as PC. Vendors of these systems have come up with what has become a
standard user interface, namely X Windows, not to be confused with Windows from
Microsoft.

The X Window standard, in simple terms, provides a method for displaying
multiple applications simultaneously on screen. At the same time, as a side benefit, it
provides one with a standard format to port applications to other systems.

This system was developed at the Massachusetts Institute of Technology
(MIT) with assistance and support of Digital and IBM. X Window is a network
transparent windowing system. This means that X Window is not tied to any given
network protocols. In turn, this means that you can run applications on any machine
on your network and control them from individual windows on your terminal. These
applications doni even have to be made by the same manufacturer or be running on
the same operating system. They just have to be in your network and running X
Window. X Window is basically software independent at the user’s graphical display
end.

This new graphics interface promises to bring the look and feel that is
common to all workstations. This means pull-down menus, dialog boxes, icons, and
other elements that the user can come to expect from every application.

9

JUSTIFYING AND PURCHASING

As workstations become the platform of choice, the abundance of ways to buy a
CAE system is making the selection process more complicated, and justifying the cost
for an entire CAE system is not easy. This technology is very new, and as such, it
produces a level of uncertainty and mistrust in management levels of most en¬
gineering organizations.

In the government, as in most engineering organizations, the promise of an
automated solution also seems and feels like an unlikely proposition.

The requirements in a military scenario are, although very related to private
industry, a bit more demanding and so are the CAE systems to be justified for pur¬
chase.

System Costs

CAE has costs that are hidden and those that are obvious. The sticker price of
hardware and software for a modest system is in the tens of thousands of dollars. Less
obvious are expenses for maintenance, training, and even loss of productivity during
the transition period to an automated system.

Initial acquisition costs very widely depending primarily in the software that one
intends to use for a given application. The hardware part, most commonly known as
the platform, is usually not as great as its software counterpart; in part due to the fierce
competition in the market for such products and because of mass production of key
components iiku memory chips and newer generation microprocessors that are
becoming common in the market place.

Operating Costs

Annual operating costs tent to follow initial acquisition costs. In a study of
300 firms, conducted in 1990 by Practice Management System Ltd., Newton, MA, it
was found that median annual operating costs were $47,775. These costs included
hardware and software, maintenance and upgrades, system operating costs of
training, and other costs (space, utilities, insurance, and supplies).

it has to be pointed out that the costs of a CAE system are specially sig¬
nificant when considering that the average life of hardware is 24 to 30 months. This
means planning ahead for replacing a updating obsolete and outdated systems.

10

User Training

This in an important but often neglected element in the purchase decision.
Organizations often succeed in selecting a good system from a reputable vendor at a
fair price, but fall short in budgeting for appropriate training. To use CAE as an
effective engineering tool, proper training available from different sources is a must if a
CAE system is going to be something more than a “dust collector.

Vendors of the software to be used are usually the best qualified to provide
basic and advanced training; although there are training companies that can provide
training for almost all the software available in the market. Some software vendors
have on-going training on advanced concepts that are usually provided for an extra
fee.

Upper management should realize that one or two training courses will not
produce immediate CAD/CAM/CAE experts but will, however, give employee’s the
confidence to go on learning the technology.

Real Benefits

The benefits of a CAE package are more difficult to quantify that costs. One can
easily say that such a system decreases the time to design or troubleshoot a circuit,
but it is more difficult to say by how much and what cost savings that implies, or how
much return the investment is paying-off. The payback of CAE can be, and usually is,
quantified by how well it increases productivity or how significant it reduces the
product development cycle. Processing engineering change proposals (ECP) is one
way in which engineering groups have already reaped benefits from their investment.

it should be understood that high productivity gains are not going to be ac¬
complished by acquiring expensive systems and using them for drafting or just simple
ECP processing, it is in design, troubleshooting, and in circuit optimization that a CAE
package gives you tangible benefits.

Another well defined benefit is the shortening of the development cycle. Com¬
monly, all the different phases of development are performed independently and
usually involve some degree of effort and resource duplication. A CAE system can
take the task of putting all the pieces together in one data base. This would shorten
the time to market of a commercial product or would cut the time of first fielding in the
military. It is indeed this ability that makes a CAE system a vital tool in all engineering
organizations.

11

Attention should be focused on the early stages of the design. Design for
performance, manufacturability/producibility. and concurrent engineering are techni¬
ques that attempt to change CAE into a system that describes costs, features,
specifications, design intent, and ranges of performances. Also, beyond accelerated
aging models, there are attempts to simulate the deteriorating effects of dormant
storage; that is a major concern in the Military.

The Purchase Decision

Seiecting a Vendor

Standardization of hardware platforms has benefitted users who now have
more choices of CAE systems. Most companies and organizations begin the process
by forming a CAE acquisition committee which requests proposals from various
vendors, and would sometimes bring them in or go out for demonstrations of the
packages that seem to fit the intended applications.

Vendors would then come back with questions relating to the company’s
needs and practices. Some of the questions may include:

• How many users (engineers, technicians, analysts, drafters,
programmers) will be using the system?

• Are you a research, industrial, or a military concern?

• Do you have an Investment in existing hardware and software that
will need to be taken into consideration.

• Are other corporate or organizational divisions using CAE equi¬
pment? If so, what kind?

• What is done more of: electronic analysis, printed circuit board
layout? design integrated circuits, millimeter wave, application specific 1C (ASIC), etc.?
statistical, or other specialized simulation as temperature analysis, or VH application
specific 1C (VHSIC) and hardware description language (VHDL)?

• Are you interested in analog or digital simulation? both? mixed¬
mode?

• How much time or what percentage of time is spent generating new
drawings? revising existing drawings?

• What is your procedure for generating bills of materials or parts
lists? technical illustrations? bids?

• Finally, and invariably, how many dollars are to be invested?

12

This and other pertinent information heips vendors understand the
operation, and makes it possible for them to make recommendations on the optimum
system configuration for each particular need.

Evaluating Vendors

It is important to buy a system that is fast enough to handle the needs of the
intended applications, and the relative speeds of the hardware to be acquired will
probably have to be discussed using ‘benchmarks’ as figures-of-merit.

Benchmarking Systems. Benchmark tests are “canned” programs that are
used for measuring computer speeds. These may or may not demand the same
capabilities from a computer that the applications do; therefore, it * important not
to select a vendor based solely on favorable scores of formal bene < tests.

A well known benchmark figure is millions of instmetions per second (MIPS).
This widely used number is the rate at which the computer can move bytes of infor¬
mation around and perform simple arithmetic on them like addition, subtraction, and
multiplication. The performance of a DEC VAX 1 1/780 as 1 MIP has become a
standard, and performance in MIPS is quoted relative to it. In theory, one can obtain
this value by multiplying a processor’s average instmetion time (AIT) by its clock rate,
assuming that the processor operates in a zero-wait-state (given that the processor
does not have to wait for data to be fetched from memory and that it has all the
information it needs readily available for its use) environment. Another often quoted
benchmark figure is the Unpack, which measures floating-point performance by
having the computer solve a 100 x 100 algebra matrix. If the application requires
mostly integer manipulation, a fast Unpack rating is meaningless; hence, the only tmly
useful evaluation of a computer is to put it through its paces processing intended
applications.

CONCLUSIONS

If one thing can be said about a computer aided engineering (CAE) system, it is
that it had thus far brought results to the great majority of users who ventured resour¬
ces in CAE. Companies of all sizes report increased productivity, reduced design and
manufacturing costs, and work of better quality. CAE tools are the fastest growing
applied technology in the world. With combined sales of hardware and software
projected to reach the 10 billion dollar mark by 1994 in the United States alone, CAE
and design is one of the leading causes of profit reinvestment in industry, and a major
source of new white-collar employment.

It is anticipated that in the immediate future hardware will be sold directly by
computer vendors or their dealers. The most popular operating systems will be the
ones available for multiple platforms. Specialized applications software will be sold by
its developers, and some may be sold by computer vendors.

13

What’s in store for CAE in the 90s? More changes. For instance, ‘open’ systems
will dominate making fiies easily accessible regardless of platform. The newly
implemented computer aided design (CAD) framework initiative (CFI) expects to
define a fuiiy integrated electronic design automation (EDA) environment. In this
environment, toois from different vendors will work together. Every tool will have the
same look and feel. Although CFI appears free of problems, it still has some critical
aspects to work out. When the electronic design interchange format (EDIF) was
demonstrated at the 1988 design automation conference (DAC), severai vendors
proved it was possible to pass design information from one vendor’s tool to another.
All the designer needs it an EDIF writer to translate design information from one EDA
vendor’s format into EDIF, and an EDIF reader to translate design information into the
EDA vendor's format. Problems arise when vendors support reading, but not writing
EDIF. In other words, vendors support the flow of information into their environments
from other’s vendors’ tools, but users can not transfer design information out to use
with competitor’s toois. At this point in time there are more EDA companies supporting
only and EDIF reader, and not an EDIF writer.

The affordable price/performance ratio of workstations and personal computers
(PC) has brought new users into the market at ail levels and users in the industry as
weil as in the Government, have a wide selection of platforms, operating systems, and
engineering software.

The electronics industry is not, by any means, the only one that has benefitted
from the CAE technology; as a matter of fact, virtually 100% of all industries have been
touched or influenced, for the better, by the CAE revolution. In the mechanical and
electronics industry the applications are even more infiuential. to the point that none of
these businesses couid survive and thrive in the highly competitive industrialized
world without the aid of CAE.

14

DISTRIBUTION UST

Commander

Armament Research. Development and Engineering Center
U.S. Army Armament, Munitions and Chemical Command
ATTN: SMCAR-IMI-I (3)

SMCAR-FSP>E, H. Neira (10)

SMCAR-GCL

Picatinny Arsenal. NJ 07806-5000
Administrator

Defense Techni^l Information Center
ATTN: Accessions Division (12)

Cameron Station
Alexandria. VA 22304-6145

Director

U.S. Army Material Systems Analysis Activity
ATTN: AMXSY-MP

Aberdeen Proving Ground, MD 21005-5066
Commander

Chemical/Biological Defense Agency

U.S. Army Armament. Munitions and Chemical Command

ATTN: AMSCB-CII, Ubrary

Aberdeen Proving Ground, MD 21010-5423

Director

U.S. Army Edgewood Research, Development and Engineering Center
ATTN: SCBRD-RTB (Aerodynamics Technology Team)

Aberdeen Proving Ground, MD 21010-5423

Director

U.S. Army Research Laboratory

ATTN: AMSRL-OP-CI-B, Technical Library

Aberdeen Proving Ground, MD 21005-5066

Chief

Benet Weapons Laboratory. CCAC

Armament Research, Development and Engineering Center

U.S. Army Armament, Munitions and Chemical Command

ATTN: SMCAR-CCB-TL

Watervllet, NY 12189-5000

Director

U.S. Army TRADOC Analysis Command-WSMR

ATTN: ATRC-WSS-R

White Sands Missile Range, NM 88002

15